Enterprise versus "consumer" grade drives

[miso]

Basically the caps have ESL (equivalent series inductance) as well as
ESR. In the case of electrolytics, it is a bit more complicated than
that. Generally if you want to cover your bases, you use a combination
of electrolytic and ceramic, but cranking the switching frequency up to
a MHz allows for ceramics without the electrolytic in many cases.
Basically the current is dumped from the inductor "fast enough" that you
don't need a large reservoir cap. The tradeoff is more loss in the
inductor core due to the higher frequency, and more difficulty in
meeting EMI specifications.

The SMPS filter requirements are also a function of how the current is
maintained in the inductor. If the current is completely dumped, this is
known as discontinuous conduction. Generally an easier supply to design,
but more ripple. If you maintain current in the inductor at all times,
then the capacitor requirements are greatly reduced, but the SMPS is
more difficult to control. This is known as continuous conduction.

 

[miso]

Incidentally, there is an hilarious (from an electrical engineering
viewpoint) paragraph in that Steve Jobs biography where he claims that
Apple invented the switch mode power supply. Christ, the bull shit that
came out of Jobs' mouth at times is simply unbelievable.

http://www.arcfn.com/2012/02/apple-didnt-revolutionize-power.html

I don't know all the history covered in that article, but Robert
Boschert clearly made switchers a commodity item. If anyone deserves
credit, it would be him.

 

[Man-wai Chang]

If you use RAID0, your data is basically doomed anyways, unless

you have good backup. If you have good backup, no need to go for
enterprise drives. RAID0 is basically always a bad choice except
for cache and buffer applications that need high throughput, such
as video-capture and editing. But RAID0 should never be used as
actual longer-term storage.

You always need to have a backup, even when you are using RAID 99.9!

--
@~@ Remain silent. Nothing from soldiers and magicians is real!
/ v \ Simplicity is Beauty! May the Force and farces be with you!
/( _ )\ (Fedora 17 i686) Linux 3.6.1-1.fc17.i686
^ ^ 18:39:02 up 4 days 10:40 0 users load average: 0.00 0.01 0.05
ä¸å€Ÿè²¸! ä¸è©é¨™! ä¸æ´äº¤! ä¸æ‰“交! ä¸æ‰“劫! ä¸è‡ªæ®º! è«‹è€ƒæ…®ç¶œæ´ (CSSA):
http://www.swd.gov.hk/tc/index/site_pubsvc/page_socsecu/sub_addressesa

 

[Rod Speed]

Grant wrote

Are "bad caps" always electrolytics

Nope, anything can fail. That isn't whats generally
meant by the term bad cap with motherboards etc tho.

and meant for filtering ripple current?

Usually but not always, particularly in switch mode power supplys.

In such a scenario, I can see the value in higher capacitance, but is that
usually the case?

The same capacitance is usually fine.

 

[Grant]

Don't use higher values! Switching regulators need to be carefully
tuned to voltage drop, frequency, inductor and capacitor values.

I've not seen it in the dozens of successful repair jobs. I do understand
the effect of poor switcher design that is dependent on output filters,
but such poor designs should not reach mass produced items, unless it is
one where many components were not fitted due to bad production engineering.

In the worst case, higher output filter caps can cause oscillations
and destroy your curcuit. I have seen oscillations as low as 0.25Hz
from this effect before I understood it. (I once build a PSU for a
computer myself, fed from a 200W transformer at 24V. That adventure
took a significant part of my spare time for 3 months, but I learned
a lot and it did work reliably at the end.)

Poor loop response to transients, easy to design around.

And yes, "bad caps" are low-ESR capacitors used as output filter
capacitor in switching regulators. But the capacitor is only half
of the filter, the other half is the inductor. LC circuits are
tricky. They do have a main resonance frequency and more at
multiples of that main one. Switching regulators have to be
carefully designed to deal with that, do not change either L or
C values on an existing design without very good reasons.

Sounds like you hit the minimum load requirement? A switcher finds
it very difficult to wind back to zero output current. Sometimes
minimum load is stated on the power supply label.

Grant.

 

[Grant]

"miso" commented:

[ ... ]

Tantalum caps can also fail.

Lots of cap videos on the net.

Surface mount caps are usually what people end up exploding by accident, especially a certain type
of capacitor where the polarity is determined by a little nib sticking out. They literally jump
off the PCB and land on the bench still smoldering if not flaming.

http://www.newark.com/vishay-spragu...lum-capacitor/dp/53K6051&ICID=Vishay_Passives

This is a decent cap except for the violent failure mode. Due to the density of the device, these
are used where space is a premium. Also due to the low ESR, they are used in high end devices.
Just spec them conservatively.

Thanks for the background info!

In linear supply design, you reach a point where it makes more sense to actively regulate than to
brute force filter. A transistor, if maintained in a safe operating area, is very reliable. Big
ass electrolytics will fail eventually. There is a trend to increase switching frequencies in SMPS
simply to be able to use ceramic caps. SMPS in excess of a MHz isn't all that odd.

Yeah, those electrolytics seem to be the limiting factor
in MTBF calculations. In college I had a part-time job in
sales for a power supply company, and to estimate MTBF,
the electrolytic capacitors were the only components
considered.

Cheaper electrolytic caps have only 2000 hours at rated temperature

Better ones are 10,000 hours at higher rated temp, run them cool to get
really long life.

Grant.

 

[Grant]

....

I guess a little research is necessary before substituting caps.
I'll check the online forums for comments on the particular board
before substituting.

Pick ones with high ripple current at 85'C and 100kHz, you pay for
quality. The ripple current rating for 100/120Hz is only meant for
mains rectified power supply, not a switcher.

OTOH you can get really cheap caps direct from China, (eg. dx.com)
probably same type as fitted to Chinese made mobos! Your choice...

I've not yet bought the cheapies, bad caps can spike a CPU to death.

Grant.

 

[Arno]

Incidentally, there is an hilarious (from an electrical engineering
viewpoint) paragraph in that Steve Jobs biography where he claims that
Apple invented the switch mode power supply. Christ, the bull shit that
came out of Jobs' mouth at times is simply unbelievable.

Hehehe, also known as his "reality distortion field". The man
did not understand the world at all, and grossly overestimated
his own skills and insights. But his particular set of delusions
sold really well and it is no accident that Apple marketing
and fan-base has aspects of a cult, i.e. something were the
"great visionary" (high priest) is not questioned and can
only speak truth. Apple engineering is not that good, they do
have lemmons and real screwups, just like everybody else.

Job's delusions also killed him in the end when he thought he
could fight liver cancer with alternative "medicine". That
mistake he did realize before he died, brobably due to the
reality distortion field failing under extreme stress.

I don't know all the history covered in that article, but Robert
Boschert clearly made switchers a commodity item. If anyone deserves
credit, it would be him.

And if you go a bit farther, it is just an on-off regulator with
an LC filter and a tiny quirk for efficiency (the diode). Clearly
a case of natural development, i.e. as soon as the basic components
were available with sufficient quality, it was bound to be discovered
by a lot of competent people.

Arno

 

[Arno]

Don't use higher values! Switching regulators need to be carefully
tuned to voltage drop, frequency, inductor and capacitor values.

[/QUOTE]

I've not seen it in the dozens of successful repair jobs. I do understand
the effect of poor switcher design that is dependent on output filters,
but such poor designs should not reach mass produced items, unless it is
one where many components were not fitted due to bad production engineering.

I have seen it in one such job (my own). It did not kill the board,
but it made it unreliable. Reduced the caps back to original values,
problem gone, ran for another 2 years 24/7.

Poor loop response to transients, easy to design around.

Poor loop response, period. Unless you count power-up as a
transient. Reducing C by a factor of 10 fixed it. At that time
I had no clue about switcher design and though increasing
the C values from the reference design would increase output
stability. Not so! Later I found that quite a few professional
PSUs also state a maximum capacitive load. Makes a lot of sense
once you understand the details.

Sounds like you hit the minimum load requirement? A switcher finds
it very difficult to wind back to zero output current. Sometimes
minimum load is stated on the power supply label.

No. This was for a CPU regulator before CPUs went into low
power states. I just managed to hit a resonance because I
increased the capacitor values significantly. I possibly
could have fixed this also by increasing them even more.
But the lesson is: Do not change the original capacity
values. You can use better quality all you like though
and of course only the whole filter bank needs to have the
same combined value, not each indicidual capacitor.

Arno

 

[miso]

And if you go a bit farther, it is just an on-off regulator with
an LC filter and a tiny quirk for efficiency (the diode). Clearly
a case of natural development, i.e. as soon as the basic components
were available with sufficient quality, it was bound to be discovered
by a lot of competent people.

Arno

Kind of like the notion of synchronous rectification. Until cheap power
mosfets came along, the whole idea seemed crazy. Now it is done to be
efficient.

I see Fry's is starting to put the 3TB Seagate on "special". They had
the 4tB deskstars on sale about two weeks ago, but they were all
returns. Some software/systems can't handle the large drives.

This is the old advert:

http://www.frys-electronics-ads.com...B-DeskStar-Serial-ATA-600-Internal-Hard-Drive

You can always track down old adverts from this website. For instance,
to see all 4tbyte drives,

http://www.frys-electronics-ads.com/tag/4TB+hard+drive

Change 4 to 3 to get 3Tbytes, blah blah blah.

 

[miso]

I wouldn't change the value of the cap. Lower ESR than the original is
probably OK if the parts can take the ripple, but don't change the value.

 

[miso]

I use a double conversion UPS, so I am generally more worried about
power than most people. Cost of power matters too of course, but with
double conversion UPS, you have a hard limit.

If you have an old P3 running 24&7, at some point it makes sense to look
at Atoms. Given that intel hasn't put Atoms on the 22nm line, it is
getting to the point where some of the LV Xeon's are getting close to
Atom territory. Intel has a 17w TDP Xeon. E3-1220LV2 Of course the mobos
are more complicated than Atom, so the overall power is higher.

It is hard to fight C*V^2*F. At some point, the finer geometry will beat
all your head banging power saving designs used in a larger geometry
process.

 

[Arno]

Kind of like the notion of synchronous rectification. Until cheap power
mosfets came along, the whole idea seemed crazy. Now it is done to be
efficient.

Indeed. And once the Mosfets were available, it was obvious to do it
that way.

Arno